> If, through grade inflation, the only thing that matters to an employer is what school you went to and that you completed it (the sheepskin effect), then isn't the correct optimization to reduce wasting time on levers that won't make any practical difference in the end?

The reason I don't think this is rational at all is the amount of work needed to "look good for employers" isn't really that far off from the amount of work needed to understand and learn the info well in the first place.

I used to do a lot of college hiring for software devs. We did on-campus recruiting at a bunch of top universities, so sure, the school you went to is inherently one factor in our hiring process. But we also definitely cared about the grades you got, especially in core CS courses. Most importantly, my on-campus interviews were focused on things that someone should have learned in their data structures and/or algorithms course (but used examples that were as "real world" as possible). If you didn't actually understand the material, we weren't going to hire you.